Re: GCC build of HEAD failed for native with your patch on 2004-09-07T00:05:16Z.

[Devang Patel <dpatel at apple dot com>]

On Sep 6, 2004, at 8:10 PM, James E Wilson wrote:

> On Mon, 2004-09-06 at 18:19, GCC regression checker wrote:
> > With your recent patch, GCC HEAD does not compile on:
> >  native
> > Attached is build output for those targets.
>
> > ______________________________________________________________________
> In postreload.i.04.cse we have
> (insn 1364 1363 1366 151 (set (reg:SI 174 [ T.882 ])
>         (neg:SI (reg/f:SI 732))) 75 {negsi2} (nil)
>     (expr_list:REG_EQUAL (const:SI (minus:SI (const_int -360
> [0xfffffffffffffe98])
>                 (symbol_ref:SI ("recog_data") [flags 0xc0] <var_decl
> 0xc24c1c recog_data>)))
>         (nil)))
>
> This is a rather odd construct, as the negative of an address, isn't
> really useful for much.  Makes one wonder if maybe something has 
> already
> gone wrong earlier.
>
> This odd construct confuses rs6000_machopic_legitimize_pic_address 
> which
> only handles (const (plus ...)).


I think, Jan Hubicka addressing this.

	http://gcc.gnu.org/ml/gcc-patches/2004-09/msg00584.html

Later on in the thread, he has posted revision of his patch also.

-
Devang